AI Paper Atlas

Research map / Continual learning / Continual fine-tuning of LLMs

MagMax: Leveraging Model Merging for Seamless Continual Learning

Daniel Marczak, Bartlomiej Twardowski, Tomasz Trzcinski, Sebastian Cygert

ECCV 2024 · Continual learning › Continual fine-tuning of LLMs · approach: Low-rank and model-merging adaptation · 37 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use sequential instead of independent fine-tuning before maximum magnitude merging (ablation)MagMax · Class-incremental average (14 settings)task-agnostic accuracy (%): 71.17 -> 74.50Improved
use sequential fine-tuning with TIES merging (ablation)TIES · Class-incremental average (14 settings)task-agnostic accuracy (%): 71.84 -> 74.37Improved
use sequential fine-tuning with Avg merging (ablation)Avg · Class-incremental average (14 settings)task-agnostic accuracy (%): 71.75 -> 73.05Improved
use sequential fine-tuning with Avg merging (ablation)Avg · Eight-dataset benchmarkaverage accuracy (%): 67.54 -> 79.65Improved
use sequential fine-tuning with MagMax on dissimilar tasks (ablation)MagMax · Eight-dataset benchmarkaverage accuracy (%): 74.14 -> 82.74Improved
add maximum magnitude merging to LwFLwF · Class-incremental average (14 settings)task-agnostic accuracy (%): 72.36 -> 73.85Improved
add maximum magnitude merging to EWCEWC · Class-incremental average (14 settings)task-agnostic accuracy (%): 70.79 -> 73.97Improved
start fine-tuning from first-task model theta_1 instead of pre-trained weights (ablation)MagMax · Class-incremental average (14 settings)task-agnostic accuracy (%): 71.17 -> 70.37Hurt
start from previous checkpoint theta_{t-1} instead of theta_1 (ablation)MagMax · Class-incremental average (14 settings)task-agnostic accuracy (%): 70.37 -> 74.50Improved
merge by maximum magnitude versus lower-magnitude selection for sequentially fine-tuned modelsMagMax · CIFAR100 magnitude analysisaccuracyImproved
use simple averaging or random mixing of independent fine-tunings instead of sophisticated mergingmodel merging baselines · Class-incremental average (14 settings)task-agnostic accuracy (%): 71.75 / 71.78 vs TIES 71.84No clear effect
apply sequential fine-tuning to RandMix on fine-grained and Cars/20 settings (ablation)RandMix · Cars /20task-agnostic accuracy (%): 64.95 -> 62.18Hurt

Later papers that compared with this method

4 later accepted papers report MagMax as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy